Perplexity AI received another huge amount of financing, but at the same time it was caught in a “plagiarism controversy”

According to the latest report from Bloomberg, SoftBank Group’s Vision Fund No. 2 is about to invest in the American artificial intelligence startup Perplexity AI.

According to people familiar with the matter, SoftBank’s investment amount is US$10 million to US$20 million, and Perplexity’s total financing in this round has exceeded US$250 million.

The funding round is expected to triple Perplexity's valuation to potentially $2.5 billion to $3 billion, making it one of the most highly valued companies in the industry.

Perplexity aims to use artificial intelligence to compete with Google search.

As a start-up unicorn company, their core service is to provide an "answer engine", which is fundamentally different from traditional search engines.

Instead of searching through multiple results to find the primary source of your question, users get the answer directly that Perplexity finds for you.

Aravind Srinivas was a research scientist at OpenAI. After leaving OpenAI, he founded Perplexity in August 2022.

Perplexity wants to provide users with fast, accurate answers without requiring them to sift through mountains of information.

Aravind Srinivas also said in an interview with The Verge:

We care about authenticity and accuracy.

As "the world's first conversational answer engine", Perplexity's answer interface is very clean. On the results page, there are sources of information at the top, answers in the middle, and extension questions below.

Its unique feature is that it combines ChatGPT-style question and answer with the link list of traditional search engines to create a new search experience.

In an earlier interview with Wired, Huang said that he "has been using Perplexity."

Of course, he also thinks ChatGPT is very good. During the interview, Huang Renxun was particularly interested in the field of computer-aided drug discovery. He "used these two almost every day" for research:

Perhaps you want to learn about advances in the field of computer-aided drug discovery.
Then you have to build a framework around the topic first, and then ask more specific questions from that framework.

While Perplexity's business model is attractive in theory, its role as a middleman may cause some content creators to worry.

Like Arc Search and Google Gemini, Perplexity provides answer results directly after searching for a question.

If you do this, it will definitely affect the traffic and advertising revenue of the original content website.

For traditional search engines like Google and Baidu, most of their information is automatically crawled by crawlers to facilitate users to search through keywords.

A crawler can quickly retrieve and organize a website's information, but it won't crawl all content mindlessly. When a website is generally created, a Robots protocol file (i.e. robots.txt) will be set.

Through this file, the website can tell the search engine crawlers: which web pages can be crawled and which cannot. This is a non-mandatory agreement that relies primarily on compliance by crawler developers.

Most search engines and crawler developers will respect the Robots Agreement and will not crawl content that the website explicitly prohibits crawling. This is done to respect the website's privacy and copyright, and to avoid legal issues.

If this protocol is not followed, the crawler will force access to the website's content. Another consequence of this is that some website paywalls may become ineffective.

Not long ago, someone was using Perplexity and tried to get it to summarize Eric Schmidt's secret drone project.

However, in the results given by Perplexity, you can see that several clips are taken from Forbes' exclusive reports, and an original illustration created by Forbes is also given.

Later, Forbes also actively tried to use Perplexity to generate content. After searching the topics of multiple articles, they found that their own text content often appeared in the answers given by Perplexity.

Among them, many articles are exclusive reports with paywalls. Perplexity bypasses Forbes' paywall and uses many original images and text without permission.

Not only that, but the source of the information does not appear in the text of the answer given by Perplexity. The only attribution is a few icons that link to these media, but they are very small and easy to miss

In addition to Forbes, another well-known media outlet Wired is also suing Perplexity.

Wired wrote an earlier article about Perplexity, which also reported that Perplexity is trying to use crawlers to crawl blocked website content.

But then, something weird happened: Perplexity "stealed" this article from Wired – even though the article was about itself, and Wired explicitly blocked Perplexity's access in the relevant documents on its website.

Wired developer Robb Knight took a look at the data behind the scenes.

After extensive analysis, Wired identified a specific IP address that had a high probability of being associated with Perplexity and was not within Perplexity's public IP range.

In an effort to calm the mood at content creation sites, Perplexity's chief commercial officer Dmitry Shevelenko said in an interview with Semafor that Perplexity is developing revenue sharing plans with publishers.

Perplexity has yet to announce details about these partners, but Dmitry Shevelenko said it would announce its plans as soon as possible.

Aravind Srinivas also responded to this issue in an interview with Fast Company:

In fact, Perplexity doesn't ignore robots.txt, it just uses a third-party crawler that ignores it.

However, Aravind Srinivas declined to name the third-party scraper, nor did he promise to ask the scraper to stop violating robots.txt.

If you try to ask Perplexity: "As an AI search engine, what do you think about citing other people's articles without authorization?"

It will give the following answer:

It seems that Perplexity itself knows that doing so is risky and illegal to a certain extent.

Take journalism for example. If you had to write a new article, what would you do?
You'd say "according to the New York Times," which is quoting someone else. This is exactly what we are doing.

So said Dmitry Shevelenko, CEO of Perplexity.

In any case, I still hope that Perplexity can continue to create more innovative AI tools while complying with regulations.

# Welcome to follow the official WeChat public account of Aifaner: Aifaner (WeChat ID: ifanr). More exciting content will be provided to you as soon as possible.

Ai Faner | Original link · View comments · Sina Weibo